KR20030015256A - Laser system for processing target material - Google Patents

Abstract

The present invention relates to an energy efficient method and system for treating a target material, such as a microrobe structure in the microscopic region, without causing undesirable changes in the electrical and / or physical properties of the material surrounding the target material.

The system includes a controller for generating a process control signal and a single generator for generating a modulated drive waveform based on the process control signal. The wavelength has a sub nanosecond rise time. The system includes a gain switched pulsed semiconductor seed laser having a first wavelength to generate a laser pulse train at a repetition rate. The drive waveform pumps the laser so that each pulse of the pulse train has a predetermined shape. Moreover, the system includes a fiber amplifier subsystem that optically amplifies the pulses to obtain amplified pulses without significantly altering the desired shape of the pulses. The amplified pulse has little distortion and substantially has a relative transient power distribution like the original pulse train from the laser. Each amplified pulse has a substantially rectangular temporal power density distribution, sharp rise time, pulse duration and fall time. The subsystem also includes a wavelength shifter in the form of an optical fiber that is controlably shifted from the first wavelength to the second large wavelength to obtain an amplified wavelength shift pulse train.

Description

LASER SYSTEM FOR PROCESSING TARGET MATERIAL}

Semiconductor devices such as memories generally have conductive links bonded to a transparent insulator such as silicon oxide supported by a main silicon substrate. During the laser processing of such a semiconductor device, energy enters the substrate or other structure while the beam is incident on the link or circuit element. Depending on the power of the beam, the length of time of application of the beam, other operating parameters, silicon substrate and / or surroundings may overheat and damage.

A review of several prior art documents discloses the neutrality of the choice of wavelength as a critical parameter for substrate damage control. US Patent 4,399, 345; 5,265,114; 5,473,624; 5,569,398 discloses the advantages of choosing a wavelength in the range of 1.2 mm or more to prevent damage to the silicon substrate.

The foregoing disclosure of the '759 patent describes the wavelength characteristics of silicon in detail. Absorption in silicon drops rapidly behind about 1 micron with an absorption edge of about 1.12 microns at greenhouse temperatures. At wavelengths above 1.12 microns, silicon begins to transfer very easily and when silicon removes material from the silicon, a good part product can be obtained. In the range of nearly 1 micron, the absorption coefficient decreases from the fourth order of magnitude, ranging from 0.9 microns to 1.2 microns. When running from 1.047 microns to a standard laser wavelength of 1.2 microns, the curve shows the second drop in magnitude. This represents a quick change in absorption for very small changes in wavelength. Thus, operating the laser at wavelengths above the absorbing edge of the substrate mitigates damage to the substrate, particularly if there is some mismatch of the laser beam to the link or if the beamed spot extends beyond the link structure. It is important. Moreover, when the substrate temperature rises during processing, the absorption curve shifts further to infrared, causing thermal runaway and significant damage.

The problem of liquid crystal repair is similar to the problem of metal link removal. The wavelength selection principle of maximizing absorption contrast is preferably applied to the green wavelength region in a manner similar to the above-described inclusion for the same purpose (removal of metal without substrate damage). The system of Florod manufacturing is described next (the publication: Xenon Laser Repairs Liquide Crystal Displays ", LASERS AND OPTRONICS, Page 39-41, April 1988).

Just as wavelength selection has proved beneficial, it has been recognized that other parameters can be adjusted to improve the laser processing window. For example, this is disclosed in the following: "Computer Simulation of Target Link Explosion in Laser Programmable Redundancy for Silicon Memory" by L.M. Scarfone and J.D. Chlipala, 1986, p. 371.

It is desirable to select laser wavelengths and different material thicknesses to enhance absorption for the link removal process and to reduce the absorption elsewhere to prevent damage to the rest of the structure. The usefulness of the thick insulating layer under the link or circuit element, and the usefulness of limiting the duration of the heating pulse, were recognized as in the applicant's paper (see "Laser Adjustment of Linear Monolithic Circuits," Litwin and Smart, 100 /. LIA, Vol. 38, ICAELO (1983).

The '795 patent discloses a tradeoff in the selection of longer wavelengths, in particular the spot size, focus depth and pulse width available from Nd: YAG lasers. These parameters are very important for laser processing at very small sizes and where there is a chance of mutual damage of the surrounding structure.

Indeed, an improvement in expanding the processing window is advantageous because the industry is driven by high density microstructures and associated geometries of either depth or 1 micron in both sizes. Tolerances in energy and target absorption are required to process microstructures in these dimensions. It is large compared to the energy required to do so. It can be seen from this paper that laser processing parameters are not required independently for micromaching applications where small laser spots of about 1 μm are required. For example, the spot size and pulse width are minimized to short wavelengths of less than 1.2 µm, but the absorption contrast is not maximized. Manufacturers of semiconductor devices generally continue to produce already developed products while developing highly improved versions of production that typically use other structural processes. While many current polysilicide or polysilicon links are used, the compact structure of the metal is used for reinforced materials such as 256 megabits. A high 1/3 micron deep and 1 micron wide link on a 0.3-0.5 micron thin silicon oxide layer is used for this large memory. The production equipment has a Q-switched diode pump laser, associated equipment capable of operating at conventional wavelengths of 1.047 μm-11.32 μm and associated equipment capable of operating in the perceived wavelength range for low absorption by silicon. However, the user realizes the advantage of equipment improvements that cause the cleaning service of the link structure without the risk of final chip damage due to conductive use or contamination near the removal area.

Still other degrees of freedom include laser pulse energy density (transferred to target) and pulse duration. It is known that pulse widths are limited to prevent damage in micromaching applications. For example, in US Pat. No. 5, 059,764, a very short pulse of 10-50 ns is generated using a Q switched laser system of a laser processing workstation. For material processing applications (such as semiconductor memory repair via link blowing and precision engraving), the output pulse width must be very short and pulse width applications of less than 50 ns are required for many applications, for example 30 ns. Appropriate choice of pulse width allows for removal (no melting).

Fast pulsed laser designs can use Q-switched or mode-locked operation. The pulse duration, standard Q-switched and pulsed laser shape can be approximated at the base level by integrating a related number equation describing the population number density and population inversion relative to the racing threshold at the pulse initiation. For the Q-switched case, for a normalized size, the larger the number of atoms of the population inverted with respect to the threshold, the faster the pulse rise time, the narrower the width, and the higher the peak energy. As the amount decreases, the pulse shape becomes wider with lower energy density.

Typically, a Q-switched laser pulse is similar to a Gaussian mixture with a Gaussian temporary distribution or exponential decay tail. As disclosed in patent '759, shorter wavelength diode pump systems can generate very short pulses, about 10 ns, when measured at half the power point (i.e., the standard fineness of the pulse period), Works. In addition to successful operation, Applicants have demonstrated the pulse breakdown characteristics provided with substantial rise time limitations, power distributions between one-half maximum points, and significant improvements in metal link blowing applications when improved using the methods and systems of the present invention. Several limitations have been found regarding the transient pulse shape of a standard diode pump Q switch laser system that includes it.

Throughout the following specification, "pulse shape" means the generation of laser pulses detected by a detector of electromagnetic radiation, where "shape" means the power between the detectors as a function of time. Moreover, "pulse width" or "pulse period" means the full width (FWHM) at half maximum unless otherwise indicated. In addition, Q-switched pulses mean a transient distribution of pulses obtained in a standard Q switched system, similar to a mixture of substantially Gaussian center lobes with very slow decay exponent tails. These waveforms are commonly referred to as "Q-switched pulse envelopes" in laser papers.

In US Pat. No. 5,208,437 (i.e., the '437 patent), examples of pulse widths of 1 ns or less are specific to memory repair applications. Initial work by the co-inventor of '437, disclosed in "Laser Cutting of Aluminum Thin Film With No Damage to Under Layers", ANNALS OF THE CIRP, Vol 28/1, 1979, as described above, a very short laser having a "Gaussian" shape. Experimental results with pulses are included. The result shown in " preferred portion of the interconnection pattern " of aluminum or the like is deleted without the layer disposed below the interconnection pattern being damaged. In practice, specifications for pulse widths of 1 ns or less with an energy density of 10 ⁶ W / cm ² are disclosed for the device. However, even if the beam is spatially formed to correspond to the interconnection pattern, there is no intervening method of the temporary pulse shape. Furthermore, Applicants 'analysis of high density memory devices with multiple layers with specified pulse widths in very fast processing approaching 100-300 ps used in the' 437 patent is not satisfactory. Overcoming this may require the fastest laser system to generate multiple pulses that process each target site, which may reduce the laser processing speed for unacceptable levels.

Experimental results that persist on ultrafast scales are published for micromachining operations. Ultrafast pulses have a period for orders of ps (10-12) at fs (10-15 sec) and contribute fundamentally to the material properties of other atoms and molecules than those found in the range of hundreds of ps-ns on a reduced scale. .

In U.S. Patent 5,656,186 and the publication "Ultrashort Laser Pulses tackle precision Maching" LASER FOCUS WORLD, August 1997, pages 101-118, the machine behavior at various wavelengths is analyzed and the machine is shorter than the spot size where the deflection of the focused beam is limited. The processed characteristic size is described.

Laser systems for ultrafast pulse generation vary in complexity and are described in US Pat. Nos. 5,920, 668, 5,400,350 and Ultrafast Lasers Escape The Lab ", PHOTONICS SPECTRA, July 1998, pp. 157-161. In order to prevent amplifier saturation followed by compression, the method generally includes a method of pulse stretching of mode lock ultrafast pulses prior to amplification, which complements the degree of micromaching and "nanomapping" of the finest possible size, The benefits of this nanomachining are provided by the machined sub-refractive index limiting, however, Applicants now present in the power available at each pulse for micromaching applications that issue unrestricted requirements for metallink blowing and multipulse. It has been found that there are practical limitations.

Applicants want to explain the principle of use of short pulses. The rapid rise times are multifaceted for a variety of reasons and are pointed out in the following paragraphs and many theoretical and empirical papers and books describe this problem. Although this principle extends to many laser processing applications where the target material is surrounded by materials having substantially different light and thermal properties, removal of the metal ring is taken as an example.

Document 1-3 below is an example.

John F. Ready, Effects of High Power Laser Radiation, ACADEMIC PRESS, NEW YORK 1971, Pages 115-116.

Sidney S. Charschan, Guide for Material Processing By Lasers, Laser Institute of America, The Paul M. Harrod Company, Baltimore MD, 1997, Pages 5-13.

3. Joseph Bernsteine, J.H. Lee, Gang Yang, Tariq A, Dahmas, Analysis of Laser Metal-Cut Energy Process Window

Metal reflectance

Metal reflectivity decreases with increasing power density of the laser pulse (reference 1). The reflectance of the metal is directly proportional to the free electron conductivity of the material. At high electric field densities when delivered by high intensity lasers, the collision time between electrons and the grating is reduced. Reducing the impulse time decreases the conductivity and hence the reflectance. For example, the reflectivity of aluminum decreases below 92% -25% when the laser wavelength power density decreases in the range of 10 ⁹ watte / cm ² . Therefore, it is advantageous to achieve high power density at the work piece in the shortest possible time to prevent the loss of laser energy for reflection.

Heat dissipation

Thermal shift during the laser pulse is proportional to the laser pulse as follows.

here,

K is the heat dissipation of the material

t is the length of the laser pulse

Thus, it can be seen that the laser pulses prevent thermal conduction both in the thermal attenuation of the substrate under the fusion link and in the material subsequent to the link.

Thermal stress and link removal

Through energy absorption, the target metal link is heated to room temperature and expands. However, the oxide surrounding this link includes an expansion material. Thus, stress builds up in this oxide. At freezing point, the pressure of the expanded metal exceeds the yield point of the oxide and the oxide cranks and the metal links explode into microparticle vapors. The theoretical crank point of the metal link occurs at the maximum stress point, which is at both the upper and lower modes of the link, as shown in FIG.

If the oxide on the link is rather thin, cranks of the oxide will only occur at the top of the link and at the oxide and this link is removed as shown in FIG. 1A. However, in the case where the oxide is rather thick, cranking can occur at the top of the link, as well as at the top, and cranking is transferred under the substrate, as shown in FIG. This is a very undesirable condition.

Q-switched laser systems can be modified to provide short pulses of various shapes. Prior art lasers that generate high peak power, ie short pulsed lasers, are standard Q-switched lasers. These lasers generate transient pulses with a gentle pulse rise time. This temporary shape can be altered using a Popel Cell pulse pulse slicer that switches out part of the cross section of the beam. In the same U.S. Patent No. 4,483,005, which is the inventor of the present invention and the assignee is the same, a number of methods for influencing (ie, reducing the laser beam pulse width) are disclosed. As described in the '005 patent, which is incorporated herein by reference, the laser pulses can be shaped to create a "non-Gaussian" shape beam by cutting off the energy outside the central lobe. It should be noted that when a very wide Q-switched waveform is converted into a narrow, uniform shape, only a small fraction of the pulse energy is used. For example, cutting off a Gaussian pulse to provide a narrow pulse with sharp rise time and flatness within 10% reduces the pulse energy by about 65%.

Similarly, in U.S. Patent 4,114,018 (i.e., the '018 patent), a transient pulse shape for generating square pulses is disclosed. 7 shows the time interval for a very flat laser power output. In this' 018 patent method, it is required to remove the temporary segment of beam intensity to generate the desired pulse.

Preferred improvements to the prior art can provide an effective way of generating short pulses with high energy enclosures within pulse periods with rapidly decaying tails. To accomplish this, a laser technique that generates a pulse shape different from the pulse shape of the Q-switched pulse envelope is desirable. These pulses have a fast rise time, uniform energy in the central lobe and a quick collapse.

As generated by lasers other than the standard Q-switch Nd: YAG, high power density pulses accomplish this best.

These benefits are performed in a preferred manner in systems employing laser techniques that differ significantly from conventional Q-switched semiconductor diodes or lamp-pumped YAG radix.

Improvements to the prior art are desirable with methods and systems for generating standard Q-switched pulses, i.e., fast rise times, pulses that are shaped differently from pulses with very uniform and high energy concentrations in the central lobe.

The present invention relates to an energy efficient laser based method and apparatus for processing a target material. In particular, the present invention relates to the use of pulsed lasers to remove or otherwise alter portions of circuitry on semiconductor substrates, and in particular can be used in evaporation metals, polysilicides and polysilicon links for memory repair. Another application may be applied to micromachining and other repair operations where it is particularly desirable to remove or refine the microscopic structure, which often has non-uniform optical and thermal properties without damaging the surrounding area and structure. Similarly, material processing operations can be applied to other micro-semiconductor devices, such as micro electro mechanical machining. There may be medical applications such as microscopic fibers or cells using small optical fiber probes.

1 is a schematic diagram showing stress cracks only in the branded surface layer of a semiconductor caused by expanded evaporated metal;

1B is a schematic diagram of stress cracks in the upper and lower layers of a semiconductor caused by expanded evaporated metal.

1C shows a mixture of Gaussians with exponential tails called “Q-switched pulse envelopes” or pre-gas laser pulses with Gaussian enhancement.

FIG. 2 is a diagram of a preferred pulse shape of the present invention treating a metal link as compared to the same total energy Q switched.

3A and 3B illustrate a method of combining two short pulses in close proximity when generating modified pulses.

4A and 4B show the results of " pulse slicing " to improve a general pulsed pulse energy enclosure.

5 is a general block diagram of a preferred laser system for laser material processing.

6A is a schematic of one form of MOPA with a black laser distributed with a semiconductor seed laser, which is a single fiber and an optical fiber amplifier that produces the desired pulse shape.

6B is a schematic diagram of a single frequency laser and fiber optic amplifier with external boost tuning.

7 is a block diagram of another laser system of the present invention including the preferred attenuator and any shifter.

8 is a graph of the temperature at the interface between the silicon dioxide and silicon substrate of FIG. 9 as a function of the thickness of the silicon dioxide layer.

9 is a perspective view of a memory link on this substrate.

FIG. 10 is a diagram of a Gaussian laser beam focused on a small width on a focus plane that includes a metal link that emphasizes the microscopic size of the link compared to the refractive limit beam waist.

11A and 11B show the results of computer finite element simulations shown in graphs for Q-switched and square pulses where time history of stress and temperature is used for metal link processing.

12 is a schematic diagram of a system constructed in accordance with the present invention wherein the generated pulse train is wavelength shifted in a controlled manner within the optical fiber, for example, to couple to the target at the absorbing edge of silicon.

FIG. 13 is a graph illustrating the shift of the first "non-moving" wavelength relative to the second shifted wavelength in the absorption edge of silicon.

14 is a block diagram of a multi-stage fiber amplifier system amplified for the input pulse train being wavelength shifted and generating an output pulse train.

Applicants have determined that improved results can be obtained in the application of metal link blowing. For example, non-Gaussian, substantially rectangular pulse shapes have significant forgetting in metal link processing where plating insulators are present. Applicants' results show that a fast rise time at 1 ns and preferably about .5 ns provides thermal shock to the plated layer of oxide performing the link blowing process. In addition, at higher low force densities, the refractive index is reduced due to the short rising pulse. A pulse duration of about 5 ns with a substantially uniform pulse waveform connects more energy to the link, reducing the energy requirement for link cancellation. A fast overall time of about 2 ns is important to eliminate the possibility of substrate damage. Moreover, the advantage of a nearly square power density pulse is that it is best on demand, and the pulse is off when not required.

A short fast rising pulse causes the link to melt and expand before heat can diffuse down the bottom of the link. Thus, stress builds up on top of the link and promotes cranking of the upper layer without cranking down the substrate.

It is an object of the present invention to provide a simple gain switch laser system capable of generating sub-nanosecond rise time pulses with short periods of several nanoseconds and fast dropouts. The state of the art of fast pulse systems includes gain switched technology, in which a low power semiconductor seed laser is quickly and directly modulated to produce a controlled pulse shape, which is then used as a pump laser. Amplified by an amplifier, such as a cladding pump optical fiber system with a high power laser diode or diode array. Such a laser system is described in US Pat. No. 5,694,408 and PCT Application No. PCT / US98 / 42050 and is a "building block" of ultrafast chirp pulse amplifier systems, such as the system described in US Pat. No. 5,400,350. .

It is an object of the present invention to improve the laser processing method and system of the prior art, and in particular to provide a laser processing method and system which is substantially different in optical and / or thermal properties in the region near the target material.

It is an object of the present invention to provide laser pulse forming capabilities of micromaching and laser material processing applications, for example, laser ablation of links or interconnection, trimming, drilling, marking and micromaching on semiconductor substrates. The desired waveform shape is generated from a gain switch laser different from the gain switch laser of the standard Q switch system.

It is an object of the present invention to provide improvements and margins for semiconductor processing, e.g. 16-256 megabit semiconductor repair, to achieve a clean treatment of the microstructure without the risk of the factory of a modern device due to contamination or conduction residues near the removal site. do.

An object of the present invention is to provide a pulse waveform rise time of a pulse duration of about 10 nanoseconds with a short, rapid pulse breakdown such as several picoseconds, to provide laser treatment of the target structure at high power density to provide Damage due to thermal shock and dispersion is minimized.

It is an object of the present invention to achieve a high power density with high power at a workstation in a very short time, i.e. a fast rise time pulse at a wavelength suitable for laser ablation processing, thereby achieving the periphery and below the target material in semiconductor laser processing applications. By preventing damage to the structure, the processing window in semiconductor material processing applications is improved.

It is an object of the present invention to treat laser energy by treating a target location with a single laser treatment pulse having sufficient power density and fast power density rise time to provide a reduction in the refractive index of a metal target structure, such as a single metal link on a semiconductor memory. It provides a very effective coupling. Rapidly rising lazy pulses have sufficient pulse duration to efficiently heat and evaporate the material of each metal target structure using a very uniform power density during the removal period. A rapid pulse drop after the target material has evaporated prevents damage to the surrounding and plating structures.

It is an object of the present invention to provide superior performance in semiconductor metal link blowing applications compared to systems using standard Q-switched lasers, ie lasers represented by Q-switched pulse envelopes with typical pulse rise times of several nanoseconds. The laser pulse is generated to provide a substantially square pulse having a pulse duration of about 2-10 nanoseconds and a rise time of about 1 ns and preferably about .4 ns. Additionally, the pulse collapse is rapid when switched off so that only a very small fraction of the energy collapses rapidly to a very low level to avoid the possibility of damaging a given pulse period, i.e., the metal substrate or non-carbon target material. To keep. A comparison of these pulses is shown in FIG.

It is an object of the present invention to extend the processing window of a semiconductor laser removal process to provide fast and sufficient removal of a microscopic structure surrounded by a material having different optical and thermal properties. These structures are typically arranged in a way that the width and space between the structures are one micron or less and stacked deeply. The application of a short laser pulse cleanly removes the target material but prevents damage to the surrounding material caused by thermal attenuation of either of the damages to the plated substrate underneath the target material.

It is an object of the present invention to machine such that a short pulse with high energy density can be applied to control a material having substantially uniform optical and thermal properties. This pulse duration is several microseconds in the material processing range where the influential threshold is nearly proportional to the square root of the laser pulse width.

In carrying out the above and other objects of the present invention, there is provided a method based on an energy efficient laser for treating a target material having a particular size without causing undesirable changes in the electrical or physical characteristics of the material surrounding the target material. Is provided. The method includes generating a laser pulse train using a laser having a first wavelength at a repetition rate, wherein each pulse of the pulse train has a predetermined shape. The method includes optically amplifying the pulse train without significantly altering the desired shape of the pulse to obtain the next amplified pulse train. Each amplified pulse has a substantially rectangular temporal power density distribution, sharp rise time, pulse duration and fall time. The method also includes controllable shifting of the first wavelength to a second wavelength different from the first wavelength to obtain an amplified wavelength shifted pulse train. The method further includes delivering and concentrating a portion of the amplified wavelength shift pulse train to at least a spot on the target material, wherein the rise time is fast enough to efficiently couple the laser energy to the target material, and the pulse duration The silver is sufficient to process the target material and the fall time is fast enough to prevent undesirable changes to the material surrounding the target material, and the second wavelength is more efficient at connecting the laser energy to the target material than the first wavelength. to be.

The target material may be part of a semiconductor device, such as a semiconductor memory having 16-256 megabytes.

The semiconductor device may be a silicon semiconductor device in which the second wavelength may be at the absorption edge of silicon.

The minimum portion of the material surrounding the target material may be a substrate such as a semiconductor substrate.

The target material may be part of the microelectronic device.

The substantially rectangular transient power density fraction is sufficient to substantially completely remove the target material.

Preferably, the rise time is 1 nanosecond or less, very preferably 5 nanosecond or less.

In addition, the fall time is 2 nanoseconds or less.

A single amplification pulse is sufficient to process the target material.

The target material has a reflectance for the amplified pulse, and the power density of the amplified pulse is high enough to reduce the reflectance of the target material to the amplified pulse and provides an efficient coupling of laser energy to the target material.

Preferably, each pulse makes the temporary power density distribution constant within 10% of the pulse period.

The material surrounding the target material has optical properties including absorption and polarization sensitivity and heat dissipation characteristics that are different from the corresponding properties of the target material.

Preferably, the repetition rate is at least 1000 / second and each amplified pulse is at least. 1 and up to 3 micro Joules of fish energy.

Preferably, optically amplifying provides a gain of at least 20 DB.

Further, preferably, the rise time and fall time are one half of the pulse, and the peak power of each amplified pulse is substantially constant between the rise time and the fall time.

Preferably, each amplified pulse angles the tail and the method further comprises reducing the laser energy at the tail of the amplified perth to reduce the fall time of the amplified pulse while substantially maintaining the power amount of the pulse. Include.

In further carrying out the above and other objects of the present invention, an energy efficient system for treating a target material having a specified size in the microscopic range without causing undesirable alteration of the electrical or physical properties of the material surrounding the target material. This is provided. The system includes a controller for generating a process control signal and a signal generator for generating a drive waveform modulated based on the process control signal. This waveform has a sub nanosecond rise time. The system also includes a gain switched pulse seed laser having a first wavelength that generates a laser pulse train at a repetition rate. The drive waveform pumps the laser so that each pulse train has a predetermined shape. Moreover, the system includes a fiber amplifier subsystem that optically amplifies the pulse train without significantly altering the desired shape of the pulse. The subsystem includes a wavelength shifter that is capable of controlling the first wavelength to a second wavelength different from the first wavelength to obtain an amplified wavelength shift pulse train. Each amplified pulse has a substantially rectangular temporal power density distribution, sharp rise time, pulse duration and fall time. The system also includes a beam delivery and focusing subsystem that delivers and focuses at least a portion of the moderated wavelength shift pulse train to the target material.

This rise time is fast enough to effectively connect the laser energy to the target material, the pulse duration is sufficient to process the target material, and the fall time is sufficient to prevent undesirable changes to the material surrounding the target material. The second wavelength is more efficient at coupling the laser energy to the target material than the first wavelength.

The fiber amplifier subsystem includes a filter coupled to the shift to narrow (reduce the optical wavelength spray) the bandwidth of the amplified wavelength shift pulse train while providing central wavelength selectivity.

The fiber amplifier subsystem preferably includes an optical fiber and a pump, such as a high power laser diode, to pump the optical fiber.

The laser diode pump source is gain switched (pulse and directly modulated) to increase diode lifetime by switching off for extended periods of time without laser processing occurring.

Preferably, the seed laser comprises a laser diode.

The system includes an attenuator that attenuates the laser energy at the tail of the amplified pulse to reduce the fall time of the amplified pulse while substantially maintaining the amount of energy of the pulse.

The pulse duration is selected as a function of the size of the particular target material. The specified material size can be below the laser wavelength.

Preferred systems for aluminum link processing include high speed, semiconductor lasers, where the first wavelength is about 1.1 μm or less and the second wavelength is about 1.1 μm. Future materials will advance semiconductor and diode technology and fiber materials are provided at longer infrared wavelengths as well as in the visible region.

The seed laser diode may be a multi-mode diode laser or single frequency (single mode) using distributed black reflector (DBR), distributed feedback (DFB) or bottom hollow design.

The spot size generally ranges from 1 μm-4 μm.

The semiconductor device may be a microelectromechanical device.

Preferably, the laser energy attenuated in the pulse tail may be attenuated by at least 10 dB within 1.5 times the pulse duration.

In a preferred configuration of the invention, the gain switched pulse shape comprises a fast rise time pulse with a fast pulse fall time, substantially a plane at the top. The "seed" laser diode is directly modulated to generate the desired pulse shape. Optical power is increased through amplification by a fiber laser amplifier to output a power level sufficient for laser processing. The gain switched pulse at the fiber laser amplifier output is focused in the target area.

In the configuration of the present invention, it is desirable to directly modulate the "seed" diode to generate a certain gain switch square pulse and to provide low distortion amplification using fiber laser amplification to provide an output pulse level sufficient for material processing.

In another configuration, the pulsed temporal power distribution of the directly modulated seed diode is improved to compensate for distortion or to compensate for non-uniformity of the fiber amplifier or other device, i.e., "smooth" rise of the output modulator. This laser treatment pulse, which focuses on the target area, has a very relative flat during the pulse period with the desired shape, fast rise time and fast collapse.

In the configuration of the present invention, when the " seed " laser pulse is terminated, the " pulse slicing " used to attenuate the laser energy remaining in the laser processing system is improved to improve the performance of the laser processing system, thereby completing the processing. Therefore, it is desirable to prevent heating of the sensitivity structure that is not hindered by the target material. A "pulse slicing" technique is useful to attenuate tails in modified pulses or standard Q-switched pulses. This is illustrated in FIGS. 4A and 4B and provided on the vertical axis of FIG. 4B in a log scale. Performing laser sag operation, in particular metal link blowing, at a pulse rate of at least 1 KHb (1000 / sec) with at least .1 micro Joules of laser pulse energy, ie .1 micro Joules emitted at the output of the fiber amplifier. desirable. The fiber optical amplifier gain is at least 20 DB (1000: 1).

In the configuration of the present invention, the laser pulse is formed to have a rise time and fall time shorter than about 1/2 of the pulse period.

In the configuration of the present invention, as shown in Figs. 3A and 3B, when coupled, it is possible to generate a series of closely spaced short pulses that produce the desired pulse shape.

In the configuration of the system using the present invention, it is preferable to operate the laser at a pulse repetition rate exceeding the material processing speed and to select a processing pulse using a computer controlled optical switch, which computer is a focused laser beam for material processing. It is operably connected to the beam positioning system used for positioning.

The objects, other objects, features and advantages of the present invention will be clearly understood from the following detailed description of the best mode for carrying out the invention in connection with the accompanying drawings.

It will be appreciated by those skilled in the art that the following examples can be used for various applications of micromachining and laser processing with appropriate adjustments to parameters such as power, energy density, spot size, wavelength, pulse width, polarization and repetition rate. will be. Particular absorption for metal link blowing is described for illustrative purposes.

In the preferred embodiment of FIG. 7, the seed laser 10 and the fiber amplifier are installed on a stable platform attached to the moving system and the workpiece. It is very important for the removal link that the beam is located at a precision less than 3/10 of a micron. The timing of the laser pulses to correlate with the relative position of the target and the optical system is important because of the continuous movement required to achieve high throughput.

The laser 10 includes a high numerical aperture optic that is externally controlled and modulated by the computer 33 and the signal generator 11 and controlled via a beam refractor, eg computer 33. And deliver to the focusing subsystem 12 further comprising a galvanometer mirror controlled by the device. The system control computer 33 is operatively connected to the positioning mechanism or mobile system 20 and the signal generator 11 to properly time the pulse generation. The laser beam must be precisely controlled to produce a sharp focused beaming beam with a spot size in the range of about 1.5-4 microns at the calibration positions X, Y and Z. For this reason, one of ordinary skill in the field of beam positioning and focusing will recognize the importance of optical devices calibrated to provide precise movement control and refractive and limited performance of the laser head or target substrate. Depending on the specific laser processing application requirements, it is desirable to provide an optical system with a narrow field of view to provide refractive index limited focusing and precision X, Y movement steps for beam processing. Moreover, several combinations of mirror movements for refraction combined with transition steps are possible.

The step and repeat table 34 are used to move the wafer 22 to a location for processing each memory die 24. Those skilled in the art of beam scanning will appreciate the advantages of a mirror-based beam refraction system, but as described above, the substitution of other position mechanisms such as X and Y transition steps for movement of the substrate and / or laser head implements the present invention. Is a viable alternative for doing so. For example, the substrate positioning mechanism 34 includes very precise (less than one micron) X, Y, and Z positioning mechanisms that operate over a limited range of movement. Positioning mechanism 20 is used to transition the laser processing optical system element including the laser, fiber amplifier and focusing subsystem in a promiscuous manner. Moreover, a more detailed description of the preferred positioning system is disclosed in pending US patent application Ser. No. 09 / 156.895 (September 18, 1998) entitled “High Speed Precision Positioning Device”.

A system optical switch 13 in the form of an acousto-optic attenuator or fockle cell is located at the laser cavity above the laser output beam. Under the control of the computer 33, the beam is prevented from reaching the focusing system except when desired and when the processing beam is required to reduce the power of the laser beam to the desired power level. During the evaporation procedure, this power level can be as low as 10% of the total laser power, depending on the operating parameters of the system and process. The power level may be about 0.1% of the total laser power during the matching procedure where the laser beam matches the target structure prior to the evaporation procedure. This aco-optic device is generally preferred because of its ease of use, although the delay of the fockle cell is quite small.

In operation, the position of the wafer 22 (or target or substrate) is controlled by the computer 33. In general, the relative motion is a substantially constant speed across the memory device 24 on the silicon wafer 22, but step and repeat motion of the wafer is possible. The laser 10 is controlled by a timing signal based on the timing signal that controls the movement system. The laser 10 normally operates at a constant repetition rate and is synchronized by the system optical switch 13.

In the system block diagram of FIG. 7, the laser beam is focused on the wafer 22. In the enlarged view of FIG. 9, the laser beam is shown focused on the link element 25 of the memory circuit or device 24.

In order to handle the precise link structure, the spot size requirement is important. The spot size requirement is typically 1.4-4 microns in diameter, with peak power occurring in the middle of the spots with good agreement with the Gaussian distribution, and low power occurring at the edges. Superior beam quality requires about 1.1 times or approaching refractive limits with generally better beam quality or squared factors. Low sidelobes are desirable to prevent optical crosstalk and undesirable illumination of properties outside the target area.

The link 25 is slightly smaller than the spot size, giving precise positioning and good spot quality. The link is, for example, 1 micron wide and 1/3 micron thick. In the case illustrated here, the link is of metal and both sizes (width) and thickness are less than the laser wavelength.

Desirable Laser System

In a preferred embodiment, the laser subsystem of FIG. 5 uses a master oscillator, or power amplifier (MOPA) configuration. The system generates a laser pulse that seeds the amplifier to generate a high power short rise pulse. Seed lasers are critical for generating fast rise times and short pulses at very low energy levels. Fiber laser amplifiers and high speed infrared laser diodes having an output wavelength suitable for laser processing applications are preferred. In the case of such a system, the laser is designed to generate laser pulses of the desired shape and speed, as shown at the bottom of FIG. That is, the fast rise time squares with the fast fall time. This pulse shape then provides the desired laser material interaction of the reduction of the metal reflectance, resulting in low diffusion of energy into the device and cracking of the top oxide without damaging the bottom oxide.

The MOPA configuration is considered a very new and pulsed version of technology. Laser diodes with sub-nanosecond rise times responsive to modulation drive power are at the beginning of the fiber laser MOPA configuration as the laser diode of the gain device. This laser diode has multiple horizontal modes, and the subsystem can be configured for single mode operation and tunes to the bulk component at the output, or alternatively to fiber grating integrated in the system.

For example, the Littman-Metalf grating configuration described in the product list by New Focus Inc in an external hollow configuration is an available configuration. 6B is a schematic of a single frequency laser with external hollow tuning, including an optical fiber pumped in the cladding by a diode laser pump.

Alternative diode laser alternatives include a distributed feedback laser (DFB) and a distributed black laser (DBL) with integrated grating and waveguide structures and in some cases with external controls that allow the user to independently control gain, phase and grating filters. do. DBL configuration including the coupler 50, see Figure 6a. This provides flexible mode selection and tuning capabilities. The laser frequency can be dynamically selected in a number of configurations by adjusting the bulk components, such as gratings and / or mirrors of the external hollow or alternatively fixed wavelengths or selected modes. The range in which the diode center wavelength can be selected affects the whole from less than 1 μm to about 1.3-1.5 μm or more and the latter wavelength corresponds to the same wavelength used for optical fiber communication. In some cases, important elements for the purposes of the present invention at the laser wavelength selected for material processing are the rise time and pulse shape of the "seed" laser diode. Also contemplated by the present invention is that the seed laser wavelength has a high gain with a small sensitivity to small wavelength changes, i. To match. In the case of Ytterbium doped fibers, the gain is high around a reasonably wide wavelength band near the 1.1 μm absorption edge of silicon. Still further development of materials or integrated fiber components can extend the useful wavelength range, which provides more flexibility when in inches with fiber emission spectra, seed laser wavelengths, and target material properties. For example, on page 92 of Photonics Spectra in August 1997, the state of the art for fiber laser development over the wavelength range from 1.1 μm to 1.7 μm is recorded.

The operation of the Raman shifter is described in the above mentioned patent, 759, regarding its specific use with short pulse Q switch systems. Preferably, the device can be located at the output of the fiber system to move the output wavelength to the desired area to improve the absorption contrast. As disclosed in the above '795 patent, since the importance of pulse width and small spot size requirements for processing is recognized, the general operation of a preferred system for metal link processing is in the range of about 1.06 μm or more with 1.08 μm. .

The output of the seed laser will be amplified for laser material processing. Preferred optical fiber laser amplifiers provide a gain of about 30 db. The seed laser output is connected directly to the core of the fiber laser or with bulk optics that split the beam for fiber delivery. However, while this technique is commonly practiced by those skilled in the art of ultrafast lasers using chirped pulses, the system of the preferred embodiment is wholly less complex than such ultrafast systems. In the system of the present invention, the seed pulse is increased and no optics are required for pulse stretching and compression. The fibers used in the amplifier system are cladding pumps with diode lasers with wavelengths, for example 980 nm, which are significantly different than the dichroic mirrors in the construction of bulk optical systems and sheath lasers that allow optical isolation of the seed and pumping beams. do.

In view of cost, size and ease of matching, the preferred configuration uses a coupling configuration in which the seed laser is directly connected to the fiber amplifier. The pump laser uses coupling technology known to those skilled in fiber laser system design to inject high power diode energy, or 980 nm wavelength, into the cladding structure of rare earth Ytterbium (Yd) doped fibers.

Less distortion is an important characteristic of fiber amplifiers. The low distribution substantially matches the output laser shape with the seed laser pulse shape and further enhances the pulse edge or uniform power shape as much as possible. The optical fiber gain medium is delivered to the optical system to generate the amplifier pulse of FIG. 5 focused on the object.

Multiple fiber amplifiers can be cascaded for further gain if desired in low distortion cases. It may be desirable to provide an active optical switch or passive optical insulator to the intermediate stage of the output to suppress instantaneous emission. These techniques are known to those skilled in the art and are disclosed in US Pat. Nos. 5,400, 350 and International Publication WO 98/42050.

In either case, it may be desirable to further refine the pulse shape by reducing the "tail" using a pulse slicer added to the laser subsystem. This may be in the form of an electro-optical device such as a fockle cell or preferably a low delay acoustic-optic modulator. With this technique, the risk of damage occurs in small multiple "pulse periods" of processing pulses. For example, if the energy is reduced by 20 dB (100: 1) within 1.5 times a given pulse period, the risk of substrate damage in metal link blowing applications does not occur for practical purposes.

To be more specific, if a pulse duration of 8 ns is selected for a square pulse shape in a metal link blowing application and the energy drops from 12 ns to 20 dB, the remaining energy is lower than the edge, which can cause damage to the Si substrate. This damage is substantial at about 18 ns after application of the laser pulse.

In a preferred form of operation, a slow delay, high bandwidth pulse slicer is operated near the end of the amplifier pulse period and has a multiplication effect on the pulse tail with the minimum distortion of the center lobe. The effect of the amplifier distortion and the "turn on delay" of the modulator is compensated to some extent by changing the shape of the seed diode laser power during the pulse period. This transient pulse shape in the focused beam is compensated and gives the desired square shape.

In addition, current fiber systems operate optimally at pulse repetition rates of about 20 KHz, slightly faster than throughput. An output optical switch, for example a slow delay acoustic optical modulator, whose driver is operatively connected to a computer, selects a pulse for processing. In this method, the reliability of the fiber amplifier and thus the processing system is high. Those skilled in the art will appreciate that it is desirable to combine the pulse slicer and the output optical switch into a single module.

Fiber amplifier system with alternative wavelength shift

When the power density is very high, the tendency of the fibers used in the present invention for Raman shift is shown by experiment. In many applications, this shift is considered undesirable because the central wavelength control is reduced and becomes more complicated due to the spectral magnification of the fiber laser. For some applications, for example, coherent communication, narrow spectral width and stability are important requirements. For example, the undesirable effects of Raman shifting (and similar Stokes or Brillouin shifts) are disclosed in International Publication WO 98/42050 and the diode laser wavelength remains unshifted after amplification of the fiber laser system for material processing. .

Raman gain or amplification results from third order nonlinear interaction with the medium. In classical optics, the medium is considered linear and properly described by the preceding system model with optical properties independent of light intensity. However, after the invention of the laser, it is determined that high light intensities can cause nonlinear behavior in a medium that can alter the speed, wavelength or absorption of light.

Moreover, "photon (photon) -photon" or "wave mixing" in which light controls the light is described. The study of secondary and tertiary nonlinear optics is an active field of research that has led to many practical methods and devices such as frequency double correction (second order). Following the basic discovery of the Kerr effect, other interesting principles have been found to include Raman amplification in magnetic focusing, magnetic guiding beams (solitons), and fiber and fertilization results (tertiary). One skilled in the art of non-linear optics will have the concept, method and analysis of the implementer.

After a detailed analysis of the nonlinearity, the Ramman gain due to its own phase modulation in the medium inducing a relationship representing the Ramman gain coefficient is proportionally affected as follows.

Therefore, Raman gain G is affected by several parameters. In an optical fiber, decreasing the coupling of wavelength λ, cross-sectional area A or refractive index n increases the gain factor γ for a given length of fiber. Similarly, increasing optical power P within performance limits also increases gain. The proportionality K is a parameter of the material. Within this limit, the gain G varies exponentially along the length L fiber "hollow".

Gαexp (γL)

Exponential gain is a common feature of optical amplifiers with feedback, which can cause a raging action. In the configuration of the present invention, the lasing action is preferably provided by a gain switched high speed semiconductor seed laser. A more detailed description of Raman gain and third wave nonlinear optics can be found at:

(Note: book FUNDAMENTALS OF PHOTONICS, Wiley-Interscience pobulication, 1991, pp. 751-755)

However, the contents and applications of the transformation are described in US Pat. 5,485,480,5,473,622 and 5,917,969. Wavelength shift occurs when the pump beam propagating in combination with the signal beam has a wavelength difference equal to the energy difference corresponding to the "Raman energy" associated with the vibrational state. It is known that silica fibers doped with germanium have a high Raman gain. Similarly, the Brillouin shift corresponds to the vibrational energy difference, although substantially smaller than the Raman shift.

It is important to note the comparable differences between passive Raman amplifiers and fiber amplifiers doped with standard Yd for illustration purposes, each of which can be used in various configurations and combinations. Note that Raman amplifiers do not have active elements in the core. The core is effectively deactivated even if the cladding is pumped with a standard amplifier or laser. The refractive index is changed by improving the germanium of the core. This then changes the mode size to increase the Raman mobility and provide high gain. Thus, Raman migration occurs in fused silica media.

For the preceding initiation where the output wavelength of the fiber amplifier used for material processing coincides with the seed laser wavelength, Applicants have limited the advantage of using Raman shift in the fiber. The increase in the spectral width (wavelength spray) is large where this increase is not a large drag factor. The wavelength shift improves the coupling between the range beam and the target material, for example, by shifting the wavelength to the absorption edge of silicon (about 1.12 μm), for example, a fast rise time of about 1 ns, a number of times. It is virtually desirable to provide pulses with nanoture pulse durations and fast fall times.

In an alternative configuration of the present invention, wavelength shift can be accomplished using Ramman amplification in the fiber 90, as shown in FIG. This embodiment results in a simple configuration that uses only the controlled ram moving output beam 91 as an advantage while at the same time relieving the state of controlled shifting.

12 and 13, in a preferred embodiment of the present invention, Raman movement generated by coupling light to the vibrational transition of the medium is achieved by reducing the size of the core or by changing the refractive index causing the directionality of the Raman process. do. The laser output 97 has the wavelength λ of the seed laser or laser diode 93 moved in the fiber 90 to produce a longer wavelength λ + Δλ output beam at the fiber output. For example, laser diode 93 may be a semiconductor diode having a center wavelength at about 1.06 μm and Raman shifting may be used to shift the output for wavelength 99 of about 1.12 μm, the greenhouse temperature absorbing edge of silicon. Do. The output beam 91 is delivered to and focused by the silicon device 95 by the system 94.

In a preferred embodiment, the amplifier stage is minimized to match the requirements for gain and output power.

In one drive of the present invention, a two-stage system is used, as shown in FIG. In either configuration or function, the same or similar elements as those in FIG. 12 have the same reference numerals, but are marked with one prime mark. Yb-doped amplifiers having a considerable wide bandwidth and having fibers 90 'pumped by high power semiconductor diodes are used in combination of the preamplifier and wavelength shifting stages. The seed laser output wavelength 97 'generated by the semiconductor laser diode 93', e.g., 1.064 mu m, shifts to a higher wavelength, about 1.120 mu m, to generate a first stage output 91 '.

The output 91 'is connected to a passive stage 901 with a fairly narrow bandwidth and high Raman gain, such as fused silica fibers doped with germanium to form a Raman amplifier by a coupler. An amplified output beam 902 branched from the fiber is collected and delivered to the target area.

Filter 903, which is a distributed or interference filter, is used to narrow the bandwidth and reject the superrius wavelength.

For example, the table below shows the parameters associated with a particular laser produced by the application for use in the present invention. The peak output energy at 1.118 μm is shown near the published silicon absorbing edge of 1.12 μm.

Seed laser wavelength 1.064 μm Output energy, speed 10uJ at 10Khz Filter length 3m Output wavelength 1.1176 μm Spectral width 6-9nm M ** 2 1.05 Pulse Rise Period, Duration, Fall Time ～ 1ns, 5-15ns, 1ns Superrius Wavelength Rejection 22DB

The Raman fiber amplifier output beam is then collected by the optical system 94 to focus and to the target region 95. In the configuration of the present invention, the Raman amplifier simultaneously provides controlled Raman shifts while eliminating the need for the output modification of FIG. 7 (eg wavelength shifter) without loss of functionality.

In addition, where standard laser diode wavelengths (eg, 1.064 μm) are useful and are combined with high power diodes, the preferred range for laser arrays used to pump useful coupling wavelengths (eg, 980 nm) with very high output power. (Eg, 1.098 μm to 1.123 μm). Such bonding is desirable for treating reflective metal link structures (eg, aluminum) while preventing damage to the auxiliary structure.

Those skilled in the art will appreciate that other combinations of diode lasers and pump wavelengths may be desirable depending on the material processing requirements. This principle can be applied to shorter wavelengths in the red-blue portion of the visible spectrum. In this gas spectrum, some materials require less laser power for proper processing due to improved coupling efficiency. Also, for example, the fiber system can be used for Raman shifting at near IR with high output power and then at a frequency tripled to the UV output.

Raman migration occurs because of the high power density of fiber lasers required for the material processing applications of the present invention. Thus, the identification advantage of the present invention results from the controllable shift of the input wavelength to generate an output wavelength that matches the requirements for material processing. Generally undesirable shifts of about 50-60 nm have been previously found throughout the experiment. It can be seen that spectral broadening occurs and this method can be used in minutes of fiber laser designs, so that the filter can be used to minimize the output spectral shift to narrow the bandwidth. Such filters may be refractive grating or bulk interference filters recorded in commercially available systems. The typical tradeoff of narrowband output power, which is optical output versus application (over a broad spectrum), is dependent.

Laser System-Alternative

There are several advantages to the preferred system of seed radar and fiber amplifiers. Current modulation of a laser diode with a suitable driver can directly produce the desired gain-exchange pulse shape amplified by the fiber laser amplifier with low distortion. This method is considered to be the best and most efficient approach to practicing the present invention. However, those familiar with the field of laser pulse generation and shaping will appreciate that other less efficient approaches may be used. For example, a variant of a Q-switched system that extends wider than the disclosure of 4,483,005 can use a number of control functions to drive a fockle cell or optical switch provided that the modulator response time is fast enough to obtain a very flat pulse. Modern techniques for pulse width include modified output couplers, for example couplers that replace Nd: YAG Q-switched lasers with either GaAs in bulk or crystal form. Q-switched pulses from several picoseconds to several nanoseconds are known for passive Qswitching of Nd: YAG lasers with GaAs output couplers from OPTICAL ENGINEERINH, 38 (11), 1785-88, September 1999.

Laser Processing Steps and Results

The metal link 25 is supported to the silicon substrate 30 by a silicon dioxide insulator layer 32, which may be 0.3-0.5 microns thick, for example. Silicon dioxide extends over the link and an additional insulating layer of silicon nitride is provided over SiO ₂ . In link blowing techniques, a laser beam impinges on each link and heats it to the melting point. During heating, the metal prevents evaporation due to the boundary effect of the coating inert layer. During a short wavelength period, the laser beam must gradually heat the metal until the metal expands to break the insulator. At this point, the molten material is placed under high pressure, temporarily evaporating and cleaning with the wind through the destroyed holes.

As disclosed in the aforementioned 759 patent, even with small metal links and very small spot sizes used, it is considered that the heat basically diffuses the exponential gradient by conduction from the portion of the beam hitting the target. By using peak beam power to deliver sufficient energy for the evaporation of the link in pulses of 8 nanoseconds, preferably substantially less than this, the conduction component of the heat transfer is very thin, although the conductive component Substantially confined, the temperature rise that contributes to the absorption of the beam of silicon and the temperature rise of the silicon that contributes to conduction remain accumulator below the temperature threshold at which unacceptable silicon damage occurs.

Moreover, the above-mentioned, 759 patent discloses several important contents regarding the heat transfer characteristics of the link and the adjacent structure. The thermal model shows that a narrow pulse width of 3-10 nm, depending on the thickness of the target material, is desirable to prevent continuous damage and thermal conduction of the Si substrate to its representative size. However, other configurations adjacent to the link may affect the quality of the laser treatment results, as shown in the following unemployment results.

The benefits of gain-switched square pulse shapes are verified through computer simulation (finite component analysis). The musk for the laser used for link blowing is as follows.

▶ Laser Wavelength 1.083 Micron

Maximum laser energy 10 micro Joules

▶ Pulse Width 7ns (FWHM, Square Pulse)

Repetition rate 10KHz (70KHz laser speed)

Spatial profile Gaussian, TEM-OO, M ² = 1.0 (temporal refraction limit)

Polarization Depolarization

Pulse rise time ～ .5ns

The laser chosen was a Ytterbium, clad pumped fiber laser in a MOPA configuration using a 98 nm pump diode and 7 micron diameter single mode fiber.

Experimental results of lasers specified on recent memory devices show superior performance compared to standard Q-switched laser systems. According to this result, the short and fast rise time of the MOPA laser shows high performance.

As explained earlier, there are three reasons for this.

1. The 1.083 wavelength is long enough to prevent substrate damage with absorption of about 10 times or less occurring at 1.083 µm compared to the 1.047 µm wavelength.

2. A fast rising pulse provides thermal shock to the coating of oxide using link removal.

3. The high power density of the fast rising pulse reduces the link refractive index, which allows for efficient energy connection.

These characteristics are very different from the interactions observed with the QSwitch system. Moreover, computer finite component models are used to simulate the effect of fast rising pulses on various material thicknesses and link sizes. The results of the computer model generated by Bernstein, the author of reference 3, are shown in FIGS. 11A and 11B. The following Tables A and B relate to the graphs of FIGS. 11A and 11B, respectively.

Table A

Model 1 @ 0.7uJ Square pulse Slow rising pulse First crack 929K @ 1.88ns 978K @ 2.40ns Second crack 1180K @ 2.93ns 1380K @ 3.45ns Third crack 1400K @ 2.05ns none Fourth crack 1520K @ 4.73ns none

A1 thickness: 0.8 μm A1 width: 0.8 μm

SiO ₂ : 0.1 μm Si ₃ N ₄ : 0.4 μm

Laser Energy: 0.7uJ

Table 2

Model 2 @ 0.7uJ Square pulse Slow rise time First crack 974K @ 2.03ns 1050K @ 2.55ns Second crack none none Third crack none none Fourth crack none none

A1 thickness: 0.8 μm A1 width: 0.8 μm

SiO ₂ : 0.6 μm Si ₃ N ₄ : 0.6 μm

Laser Energy: 0.7uJ

The stress and temperature history clearly demonstrates the importance of fast rising pulses with sub nanosecond rise times. It can also be seen that providing significant nanoseconds, e.g., 15 ns after removal is complete if significant energy is present, may damage the Si. Fast fall times are also important with high extinction.

According to the present invention, the silicon substrate is considerably cooled by limiting the pulser to rectangular pulp with rapid collapse and by the proper choice of wavelength. In this example the laser wavelength is slightly lower than the silicon greenhouse absorbing edge of (about 1.1 μm). Although the results reported here do not indicate substrate damage, it should be noted that improved margins are useful when desired. For example, Raman shifters can be used to shift the output wavelength above the absorbing edge. Alternatively, another diode laser is potentially commercially available for MOPA configurations. This wavelength selection and transfer technique is preferably used for lazy processing and micromaching applications. Therefore, by limiting the heat, it can be seen that the silicon is in a thermal runaway state where silicon damage may occur without moving the absorbing edge to the infrared.

Certain embodiments of a fast pulse generation MOPA configuration that cleanly blows metal links are considered examples of pulse formation and are provided for illustration rather than limitation. Directional modulation of the seed laser is found to maintain efficient sub-nanosecond control over the pulse shape, including the possibility of fast compensation to correct the output pulse shape. Other applications such as micromachining, marking, scribing, etc. may be desirable from precise and fast pulse control. For example, seed diodes are susceptible to being modulated into “smooth” waveforms or other non-switched waveforms within the surface. Likewise, due to the fast response of the laser diode, it is possible to generate sequences of variable width, short pulses in a fast order. Those skilled in the art of laser processing will understand the broad application of laser systems. The scope of the invention is indicated by the following claims and is not intended to be limiting.

Claims (84)

A laser based method for treating a target material having a specified size in a microscopic region without causing undesirable alteration of the electrical or physical properties of the material surrounding the target material, Generating each laser pulse train using a laser having a first wavelength at a repetition rate, wherein each pulse of the pulse train has a predetermined shape; Optically amplifying without significantly changing the desired shape of the pulse to obtain an amplified pulse train, each amplified pulse having a substantially rectangular temporary power density distribution, a sharp rise time, a pulse duration and a fall time; Controllably shifting the first wavelength to a second wavelength different from the first wavelength to obtain an amplified wavelength shift pulse train; Moving and focusing at least a portion of the pulse train amplified and wavelength shifted to a spot on the target material. The rise time is fast enough to efficiently connect the laser energy to the target material, the pulse duration is sufficient to process the target material, the hagging time is fast enough to prevent undesirable changes to the material surrounding the target material, And the second wavelength efficiently couples the lazy energy to the target material more efficiently than the first wavelength. The method of claim 1, And the target material comprises a microstructure. The method of claim 2, And wherein the microstructure is a conducting line. The method of claim 3, wherein The conducting line is a metal line and the pulse duration effectively heats and evaporates a specific portion of the metal line. The method of claim 1, And the target material is part of the semiconductor device. The method of claim 5, The semiconductor device is a silicon semiconductor device and the second wavelength is in an absorption engine of silicon. The method of claim 5, An energy efficiency method, wherein the semiconductor device is a semiconductor memory. The method of claim 7, wherein And wherein the memory has a density of at least 16 and up to 256 megabits. The method of claim 1, At least a portion of the material surrounding the target material is a substrate The method of claim 9, And the substrate is a semiconductor substrate. The method of claim 1, Energy efficiency method, characterized in that the target material is part of a microelectronic device. The method of claim 5, The semiconductor device is an energy efficiency method, characterized in that it is a micro electro mechanical device. The method of claim 1, The substantially rectangular transient power density distribution is sufficient to substantially completely remove the target material. The method of claim 1, Energy efficiency method, characterized in that the rise time is less than 1 nanosecond. The method of claim 14, Energy efficiency method, characterized in that the rise time is less than .5 nanoseconds. The method of claim 1, The pulse duration is less than 10 nanoseconds energy efficiency method. The method of claim 16, Energy efficiency method, characterized in that the pulse period is less than 5 nanoseconds. The method of claim 1, Energy efficiency method, characterized in that the fall time is less than 2 nanoseconds. The method of claim 1, Wherein the single amplified pulse is sufficient to process the target material. The method of claim 1, The target material has a refractive index for the amplified pulse, and the power density of the amplified pulse is high enough to reduce the refractive index of the target material for the amplified pulse to provide an efficient connection of laser energy to the target material. Energy efficiency method. The method of claim 1, Wherein each amplified pulse has a fairly uniform power density distribution over the pulse duration. The method of claim 1, Wherein each pulse has a uniform transient power density distribution within 10% of the pulses. The method of claim 1, The material surrounding the target material has optical properties and heat spreading properties different from the corresponding material of the target material. The method of claim 23, wherein And wherein the optical characteristic comprises absorption. The method of claim 23, wherein The optical characteristic comprises energy polarization sensitivity. The method of claim 1, Energy efficiency method, characterized in that the repetition rate is at least 1000 pulses / second. The method of claim 1, Wherein each amplified pulse has at least .1 and up to 3 micro Joules of energy. The method of claim 1, Optically amplifying provides at least 20 DB of gain. The method of claim 1, The rise time and fall time are less than one half of the pulse period and the peak power of each amplified pulse is substantially constant between the rise time and the fall time. The method of claim 1, Each amplified pulse has a tail and includes attenuating the laser energy of the tail of the amplified pulse to reduce the fall time of the amplified pulse while substantially maintaining the amount of power of the pulse. Way. The method of claim 30, The energy attenuated in the tail is attenuated by at least 20 dB within 1.5 times the pulse duration. The method of claim 1, The pulse duration is a function of a specific magnitude. The method of claim 1, Energy efficiency method, characterized in that the specified size is below the laser wavelength. The method of claim 1, And the laser is a high speed, semiconductor laser diode. The method of claim 34, The first and second wavelengths of about 1.1 μm or less are about 1.1 μm. The method of claim 1, Energy efficiency method, characterized in that the spot is about 1㎛-4㎛ in size. The method of claim 34, The laser diode is an energy efficiency method, characterized in that the multimode diode laser. The method of claim 34, The laser diode method is an energy efficiency method characterized in that the single frequency laser diode using a distributed black reflector (DBR), distributed feedback (DFB) or an external hollow design. In an energy efficient system for treating a target material having a particular size in the microscopic area without causing significant undesirable alteration of the electrical or physical properties of the material surrounding the target material, A controller for generating a processing control signal; A signal generator for generating a modulated drive waveform based on the process control signal, the waveform having a sub nanosecond rise time; A gain switched pulse seed laser having a first wavelength and generating a laser pulse at a repetition rate; The drive waveform pumps the laser such that each pulse of the pulse train has a predetermined shape; A fiber amplifier subsystem for optically amplifying the pulse train without significantly altering the desired shape of the pulse; The subsystem includes a wavelength shifter that moves to control the first wavelength for a second wavelength that is different from the first wavelength to obtain an amplified wavelength shifted pulse train, wherein each amplified pulse is substantially square. Temporary power density distribution, sharp rise time, pulse duration and fall time; A beam delivery and focusing subsystem for delivering and focusing at least a portion of the amplified wavelength shifted pulse train to a spot of target material; The rise time is fast enough to efficiently connect the laser energy to the target material, the pulse duration is sufficient to process the target material, the fall time is fast enough to prevent undesirable changes to the material surrounding the target material. And the second wavelength couples the laser energy more efficiently to the target material than the first wavelength. The method of claim 39, The fiber amplifier subsystem includes a filter coupled to the shifter to narrow the band of the pulse train to which the amplified wavelength is shifted. The method of claim 39, And the target material comprises a microstructure. 42. The method of claim 41 wherein Energy efficient system, characterized in that the micro structure is a conducting line. The method of claim 42, The conducting line is a metal line and the pulse duration is sufficient to effectively heat and evaporate a specified portion of the metal line. The method of claim 39, An energy rate system, wherein the target material is part of a semiconductor device. The method of claim 44, The semi-solid is a silicon semiconductor device and the second wavelength is at the absorption edge of silicon. The method of claim 44, The semiconductor is an energy efficiency system, characterized in that the semiconductor memory. The method of claim 46, And the memory has a density of at least 16 and up to 256 megabits. The method of claim 39, At least a portion of the material surrounding the target material is a substrate. The method of claim 48, The energy efficiency system, characterized in that the substrate is a semiconductor substrate. The method of claim 44, The semiconductor is an energy efficient system, characterized in that the microelectromechanical device. The method of claim 39, Energy efficient system, characterized in that the target material is part of a microelectronic device. The method of claim 39, The substantially rectangular temporary power density is sufficient to substantially remove the target material substantially. The method of claim 39, Energy efficiency system, characterized in that the rise time is less than 1 nanosecond. The method of claim 53, Energy efficiency system, characterized in that the rise time is less than .5 nanoseconds. The method of claim 39, Energy efficiency system, characterized in that the pulse period is less than 10 nanoseconds. The method of claim 55, Energy efficiency system, characterized in that the pulse period is less than 5 nanoseconds. The method of claim 39, Energy efficiency system, characterized in that the fall time is less than 2 nanoseconds. The method of claim 39, A single amplified pulse is sufficient to process the target material. The method of claim 39, The target material has a reflectance for the amplified pulse, and the power density of the amplified pulse is high enough to reduce the reflectance of the target material for the amplified pulse and provide an efficient connection of laser energy to the target material. Energy efficiency system. The method of claim 39, The material surrounding the target material has an optical property and a thermal diffusion property different from the corresponding property of the target material. The method of claim 39, An energy efficiency system, characterized in that the material surrounding the target material has thermal properties that differ from the optical and corresponding properties of the target material. 62. The method of claim 61, And the optical characteristic comprises absorption. 62. The method of claim 61, And the optical characteristic comprises polarization sensitivity. The method of claim 39, And the repetition rate is at least 1000 pulses / second. The method of claim 39, Wherein each amplified pulse has at least .1 and at most 3 micro Joules of energy. The method of claim 39, Optically amplifying provides energy gain of at least 20 DB. The method of claim 39, Both rise time and fall time are less than one half of the pulse duration, and the peak power of each amplified pulse is substantially constant between rise time and fall time. The method of claim 39, The fiber amplifier comprises an optical fiber for pumping the optical fiber, the pump being distinguished from the laser laser. The method of claim 68, wherein An energy efficiency system, wherein the wavelength shifter is an optical fiber. The method of claim 69, The pump pumps the optical fiber at the third wavelength, and the wavelength difference between the first wavelength and the third wavelength corresponds to the vibration transition of the optical fiber. The method of claim 68, wherein Energy efficient system, characterized in that the pump is a high power laser diode. The method of claim 39, Seed laser is an energy efficiency system comprising a laser diode. The method of claim 39, Wherein each pulse maintains a temporary power density distribution within 10% of the pulse duration. The method of claim 39, Each amplified pulse has a tail, further comprising an attenuator for attenuating the laser energy at the tail of the amplified pulse to reduce the fall time of the amplified pulse while substantially maintaining the power amount of the pulse. Energy efficiency system. The method of claim 74, wherein An attenuator attenuates the laser energy of the tail by at least 10 dB within 1.5 times the pulse duration. The method of claim 39, An energy efficiency system, characterized in that the pulse duration is a function of a particular magnitude. The method of claim 39, Energy efficiency system, characterized in that the specific size is less than the wavelength. The method of claim 72, An energy efficiency system, wherein the first wavelength is about 1.1 μm or less and the second wavelength is about 1.1 μm. The method of claim 39, An energy efficient system, wherein the spot is about 1 μm-4 μm in size. The method of claim 72, The laser diode is an energy efficiency system, characterized in that the multimode ion laser. The method of claim 72, The laser diode is an energy efficiency system characterized in that it is a single frequency laser diode using a distributed brack reflector (DBR), distributed feedback (DFB) or an external hollow design. The method of claim 72, The pump is a gain switched laser diode. The method of claim 39, And a computer coupled to the optical switch and a computer coupled to the optical switch and a subsystem coupled to a subsystem for selecting material processing pulses of the pulse train and coupled to a control latch of the selected pulse for the target material. The method of claim 68, wherein The optical fiber is a single mode optical fiber and the pump is a pump diode.